Systemic lupus erythematosus and rheumatoid arthritis

 
13. Statistics 
 
Calculations were performed using the SPSS package (SPSS for Windows 12.0: 
SPSS Inc. Chicago, IL, USA) and the EPI Info software version 5 (EPI Info, 
Center of Disease Control, Atlanta, GA, USA). Chi-square test (
χ
2
) was used for 
testing categorical data between groups or by Fisher’s exact test when 
appropriate. Continuous data were analyzed with the student’s t-test for 
- 39 -

---

independent samples or with a non-parametric test (Mann-Whitney U test) when 
appropriate. All P values refer to two-sided test and P values 
≤
0.05 were 
considered statistically significant. The Bonferroni correction was used when 
correcting for multiple testing. Odds ratios with 95% CI intervals were 
calculated using binary logistic regression or cross tabulation with risk estimate 
in the SPSS software, or by cross tabulation using the EPI Info software.
The binary logistic regression models in paper I and II were adjusted for sex, 
with males as reference, and for age at disease onset. 
Sensitivity, specificity and likelihood values were calculated for paper IV. 
Conditional logistic regression models were used to estimate the predictive 
values of developing RA for the antibodies analysed, the 
PTPN22 
1858T 
variant, and HLA-SE gene carriage. 
For paper V, haplotypes were constructed using an accelerated EM algorithm 
similar to the partition/ligation method described in Qin et al 
192
within the 
HaploView 4.0 software.
193
The algorithm estimates highly accurate population 
frequencies of the phased haplotypes based on the maximum likelihood from 
the unphased data input. The D’- and r
2
-values were derived using the 
Haploview 4.0 software.
193